THE GOLD RUSH

On January 24th, 1848 James Marshall discovered gold at John Sutter's mill in Coloma in what would later be El Dorado County. Information about the discovery of gold took most of 1848 to disseminate, however. In March, The Californian, a San Francisco newspaper printed a story about the discovery of gold. But the streets of San Francisco did not immediately empty. In May, Sam Brannan arrived in San Francisco and began to advertise the arrival of the gold rush. A store owner at Sutter's Fort and the publisher of the California Star, Brannan stood to gain from any increase in gold mining activity. In June, an estimated 4,000 to 5,000 miners were at work in the gold district. This represented a very large share of the adult male population in California.Footnote ⁷ By the end of July, 2,000 copies of a special edition of the California Star had reached Missouri. In August, the New York Herald printed a story on the discovery of gold. And in December, President Polk confirmed the rumors in his address to Congress.

By late 1848 the whole American nation and many foreign countries knew about the California gold rush. By December 1849 the number of gold miners had risen to 40,000. The number of gold miners would peak at 100,000 in 1852.Footnote ⁸ As Figure 1 shows, gold production rose rapidly, peaking around 1853.

Many headed for California in the spring of 1849 either overland or by ship. Estimates suggest that between 5,000 and 6,000 wagons left Missouri that spring. Others took overland routes that began further south or even in Mexico. The hundreds of surviving diaries suggest that those who made the overland journey were young men, typically traveling in the company of brothers, cousins, in-laws, and friends from their town or village. Frequently they elected one of their members to be the leader and some signed compacts that bound them to travel together to California. For the period 1848–1850, lower-bound estimates of overland migration are more than 101,000.Footnote ⁹ Ships had also begun to leave New York and other cities on the Atlantic seaboard for California. The ships either took the long route around Cape Horn to California or left the passengers in Panama. In the latter case, the passengers then traveled overland across Panama and took a second ship from Panama to San Francisco. For the period 1849–1850, arrivals by sea are conservatively estimated at 75,462.

Figure 1

GOLD PRODUCTION IN CALIFORNIA OVER TIME, 1848–1900

Source: Berry, Early California Gold, pp. 74, 76, 78.

Migrating to California from within the United States, migration typically required cash outlays of $100–$300.Footnote ¹⁰ For a migrant who paid $200 in 1850, this represented more than six-months wages for an unskilled laborer.Footnote ¹¹ The cost varied depending on when and how one traveled. The overland route was generally thought to be cheaper, but was more rigorous and took four to six months to traverse. Further, it was less convenient for those located near the Atlantic coast than the sea routes via Panama or around Cape Horn. The sea route was physically less demanding and could be somewhat shorter, but this depended on a number of factors, such as the amount of time spent in Panama waiting for a ship. Migrants on both routes probably experienced higher mortality both en route and once in California than they would have had they stayed home.Footnote ¹²

Once they arrived in California, most individuals tried their hand at mining. In 1849 and 1850 all that was required was a pick or shovel and a pan to sort out the gold from the rock, sand, and other debris.Footnote ¹³ Quite quickly miners began to organize themselves to build or operate cradles or long toms, basically slanted boxes with bars on the bottom to catch the gold. The main requirement for being a miner during this early period was the ability to endure long hours of physical labor. Those who found the physical labor too difficult and those with skills—often acquired from their previous occupation—moved on to other jobs.Footnote ¹⁴ They became clerks, carpenters, merchants, doctors, lawyers, or cooks.

THEORY

Theoretical models of migration share a common structure. Migrants compare the present value of lifetime earnings for alternatives including remaining in the same location. They then move if the expected net gain is positive. In a 2000 review article, George Borjas notes that three empirically testable propositions follow directly from this framework:

Borjas discusses the related question of who will move. Depending on the setting, migration can induce either positive selection—migration of high skill workers—or negative selection—migration of low skill workers. Which will occur depends on whether the rate of return to skills in the new location is high or low.

In the California context, the first of the testable hypotheses has little content. The information environment was noisy, and individual priors regarding the economic opportunities in California may have varied, but individuals were all considering migrating to the same location: California. If we had the years in which individuals migrated, then we might be able to investigate the effect of the variation in expected payoff to migrating to California over time. The data we use to examine migration are, however, from the 1850 Census of Population, which does not contain information on when the individual arrived.

The second and third testable hypotheses are relevant, however. We would expect states with more economic opportunities to have experienced lower migration. Similarly, we would expect states with higher costs of migration to have experienced lower migration.

It is unclear whether positive selection or negative selection was at work in this setting. One constraint on negative migration was the cost of the trip. Migration models implicitly assume perfect capital markets, which would allow individuals to borrow against future earnings. In practice, many migrants had to borrow the cash necessary to make the trip from friends or family. Not every prospective migrant had friends and family who were willing or able to supply the funds.

Some positive selection may have been occurring. We know more about those who wrote diaries and letters than the many other individuals who went. In particular, we usually know their previous occupation and something about their social class. Many diarists were from middle-class backgrounds, which indicates that there could have been positive selection. The diarists themselves were, however, unlikely to have been representative of all migrants, so we are limited in the inferences we can make.

Both Robert Margo and William Carrington provide dynamic models of labor market adjustments to a large positive shock to labor demand.Footnote ¹⁶ The basic intuition is fairly straightforward. Nominal wages rise initially as the full impact of the shock is felt and then fall as more individuals migrate to the region. In both models, the existing stock of labor will initially supply additional labor through some combination of reduced leisure and reduced labor in other sectors. Over time, additional labor will be supplied through migration. Margo notes, “The comparative static path followed by real wages may be more complex, but it is likely, too, that real wages rise initially and then fall.”Footnote ¹⁷ We do not analyze the primary implication of the models, because we do not have data on wages over time. The models do, however, provide a backdrop against which to understand our main results on productivity in mining and real estate wealth.

MIGRATION

To examine where individuals were migrating from, we use data from three sources: the 1-percent public use national sample of the 1850 Census of Population, a 100-percent sample of the 1850 Census of Population for three mining counties in California, and a 100-percent sample of the 1852 Census of Population for one mining county in California.Footnote ¹⁸ A significant problem with using the 1-percent sample to understand the gold rush is its low density.Footnote ¹⁹ It includes approximately 900 people in California, roughly 600 of whom were gold miners. To address this problem, a data set was collected by the authors and undergraduate research assistants from the manuscript records of the 1850 Census of Population. A 100-percent sample was transcribed from the original handwritten records for three gold mining counties in California: El Dorado, Sutter, and Yuba Counties.Footnote ²⁰ The three counties were located in the northern part of the gold mining region, which is sometimes referred to as the “Northern Mines”; we will refer to them as the northern mining counties. This sample includes 31,345 individuals who were living in 9,892 households. Eighty percent of the individuals in these counties were gold miners. The total number of California gold miners recorded in the 1850 Census was about 60,000; so this sample represents 42 percent of all miners. This much larger sample allows us to conduct a more detailed analysis of migration and outcomes than would otherwise be possible.

Table 1 compares the characteristics of the northern mining counties, California, and the nation as a whole. The population of California and of the northern mining counties differed from the national average in several dimensions. The average age was older, in large part because there were very few children in California. Unlike the rest of the country, where 52 percent of the population was male, 94 percent of the population in California, and 97 percent of the population of the northern mining counties was male. Self-reported literacy rates were slightly higher in California. This is in keeping with the view that literate individuals were more likely to migrate.Footnote ²¹ The percentage of the population that was foreign born was higher in California than in the northern mining counties.

Table 1

SUMMARY STATISTICS IN 1850

Notes:

The observations for the sample and for the nation are from the authors' data and public use sample respectively. The observations for California are the sum of the observations for our three counties (31,345) and the observations from the 1-100 public use sample for counties in California, excluding our three counties (614). Taking into account the differences in weighting, these two samples represent the approximately 92,700 people in California.

The first question we are interested in is the distribution of birthplaces of men ages 20–40 who were in California.Footnote ²²Table 2 lists the nine states of birth with the highest percentages (>5 percent) and absolute numbers of men in California. Figure 2 shows for all states of birth the percentage of men who were in California. As expected, men who were born on or near the western frontier were heavily represented in California. In absolute numbers, very populous states such as New York and Massachusetts were also heavily represented.

Table 2

BIRTHPLACES OF MEN AGES 20–40 IN CALIFORNIA IN 1850

Notes:

All statistics are for free men ages 20–40. All counts were multiplied by 1.6 to address the 1850 undercount. The percentage of the California population that various birthplaces represent is listed in parentheses.

In addition to the states of birth of native-born men, Table 2 lists the nine countries other than the United States with the greatest numbers of men in California. Mexico had the largest number of men. This is not entirely surprising, since Mexico controlled California until 1846. Further, because of geographic proximity, many Mexican miners were able to move rapidly to the gold rush. These numbers agree fairly closely with contemporary estimates, which suggested that there were 8,000–10,000 individuals from Sonora and Lower California in California in 1849.Footnote ²³ In 1850 a memorial requesting a customs house in Los Angeles stated “at least ten thousand Sonorans pass through Los Angeles on their way to the mines each spring, generally returning to Mexico in the autumn.”Footnote ²⁴ The United Kingdom was well represented, with large numbers of individuals from England, Ireland, and Scotland. Were these three aggregated with Canada, they would have edged out Mexico for the most men in California. China had relatively few men in California in 1850, although the numbers would grow significantly over the next few years.Footnote ²⁵

Figure 2

PERCENTAGE OF MEN AGES 20–40 FROM BIRTHPLACE WHO WERE IN CALIFORNIA IN 1850

Notes: Based on Table 2. NA are states that were not included in the 1850 Census.

To examine migration paths, we draw on a 100-percent sample from the 1852 Census of Population for El Dorado County. Like the 100-percent sample from the 1850 Census, the data set was collected by the authors and undergraduate research assistants.Footnote ²⁶ The sample includes observations for 11,936 individuals in El Dorado County. The 1852 Census asked fewer questions than the 1850 Census, and so on the whole is less valuable than the 1850 Census. It did, however, ask one important question that the 1850 Census did not: Where was the individual last resident before coming to California? We use this information to create probability distributions by birthplace of the state of most recent residence prior to California. For example, 46 percent of the men who were born in New York, had migrated to California, and were captured by the 1852 Census came directly from New York. The remaining men who were born in New York had most recently resided in frontier states such as Michigan (16 percent), Illinois (10 percent), and Wisconsin (8 percent).

Table 3

LAST RESIDENCE OF MEN AGES 20–40 IN CALIFORNIA IN 1850

Notes:

All statistics are for free men ages 20–40. All counts were multiplied by 1.6 to address the 1850 undercount. The percentage of the California population that various birthplaces represent is listed in parentheses.

We can then apply the probability distributions to the birthplaces listed in the 1850 Census to impute where the men were living prior to coming to California. This procedure will be reasonably accurate as long as two things hold. First, we need for the migration patterns from individual birthplaces not to have changed too significantly between 1849/50 and 1852.Footnote ²⁷ Second, we need the migration paths conditional on birthplace for native-born men in El Dorado County in 1852 to be reasonably representative of the migration paths of the population of California as a whole. As we will see later in this section, using the imputed state of last residence instead of state of birth improves our ability to predict who migrated. This suggests that imputation using the 1852 Census captures additional information not captured by birthplace alone.

Our imputed last residence highlights the extent to which miners are being drawn from the western frontier. Table 3 lists the eight states that had lost 5 percent or more of their men ages 20–40. Many of the states overlap with the states in Table 2. These states lost both a large share of men born in the state and a large share of men who were in the state prior to the gold rush. The fact that many of these states were losing such a large share of their prime age males to the gold rush is nonetheless striking. Minnesota and Utah each lost 16 percent of their prime age men to the gold rush. The absolute numbers of men are also striking. Missouri lost nearly 7,300 prime-age men, and Illinois and New York each lost more than 6,000 men. Figure 3 shows for all states of last residence the percentage of men who were in California.

Figure 3

PERCENTAGE OF MEN AGES 20–40 FROM LAST RESIDENCE WHO WERE IN CALIFORNIA IN 1850

Notes: Based on Table 3.

In Table 4, we examine the factors associated with the migration of native-born men ages 20–40 to California. We estimate a regression of the following form

(1)

The dependent variable is whether individual i was in California (one) or not (zero). This is regressed on a vector of the individual's exogenous personal characteristics (X_i) and a vector of the characteristics of the individual's state of birth or last residence (Z_i). The personal characteristics include age, race, and literacy.Footnote ²⁸

The characteristics of the individual's state of birth or last residence include proxies for both the state labor market and the average cost of migrating from that state to California. Recall that the second prediction from the theoretical models of migration was that individuals in locations with greater economic opportunities (i.e., stronger labor markets) will be less likely to migrate. We use the average real wages of laborers in the state in 1850 (the average daily wages of a laborer without board from the 1850 Social Statistics multiplied by six and divided by the weekly cost of board from the 1850 Social Statistics) and the share of men ages 20–40 in the state who were foreign born as somewhat crude proxies for the state labor market.Footnote ²⁹ Higher shares of foreign-born men are believed to negatively affect labor market outcomes and so induce migration of native-born men. This is sometimes referred to in the literature as the foreign-born push factor.

Table 4

LOGIT OF DETERMINANTS OF PARTICIPATION IN GOLD RUSH FOR NATIVE-BORN MEN, AGES 20–40 (marginal effects reported)

* denotes statistical significance at the 10 percent level.

** denotes statistical significance at the 5 percent level.

*** denotes statistical significance at the 1 percent level.

Notes:

All regressions are weighed by the person weights (100 or 101 for the public use sample outside of California, 160 or 161.6 for the public use sample inside California, and 1.6 for El Dorado, Sutter, and Yuba Counties). The number of observations is lower in column 3 than in the previous columns, because some relatively uncommon birthplaces are not represented in the 1852 Census, and thus we do not have transition probabilities to construct the weighted average of the state temperatures, distance, and transportation. The number of observations in columns 4 and 5 are lower than in columns 2 and 3 because real wages in 1850 and share foreign born in 1850 do not exist for some states/birthplaces that are on the frontier (for example, Kansas and Nebraska). Standard errors are in parentheses. All standard errors are clustered on birthplace. Using robust standard errors did not change the levels of significance.

The third prediction from the theoretical models of migration was that individuals with higher costs of migration will be less likely to migrate. We use distance from the state to San Francisco, the availability of water transport, average state latitude and average state latitude squared as crude proxies for the cost of traveling from the state to California.Footnote ³⁰ Of the four measures, average state latitude and average state latitude squared are least obviously related to cost. What we are trying to capture here is the well-documented fact that during the nineteenth century, individuals were most likely to move East-West.Footnote ³¹ One reason individuals moved East-West was that their human capital in farming tended to transfer best to locations with a similar climate. Thus, it was less costly to migrate East-West than, for example, North-South. Most migrants to California probably did not intend to farm, although many would later do so. In this case, the propensity to move East-West may also reflect other costs such as the fact that the major overland migration routes ran through Missouri and the major water migration routes left from New York. Thus, the costs of reaching these routes would be lowest for those in middle-tier states. All regressions have standard errors clustered on birthplace and are weighted to adjust for differences between our 100-percent sample and the public use 1-percent sample.

What we find in column 1 is that individual characteristics explain a relatively small fraction of the variance. The coefficients on age, age squared, white, and literate are, however, all statistically significant and of the expected sign. Individuals who were literate and in their mid-twenties were more likely to be in California than individuals who were illiterate or in the younger or older parts of the age distribution.

In columns 2 and 3, when we include characteristics of the individual's state of birth, the fit improves relative to column 1. In column 2, the coefficient on distance is negative and significant, indicating that individuals were more likely to migrate from closer birthplaces. This confirms what we observed in Figure 2. The coefficient on transport is positive and significant, indicating that individuals were migrating from states where there was greater access to water transportation. The coefficients on latitude and latitude squared are both statistically significant. They indicate that individuals are most likely to migrate from latitudes of 37 degrees, which is roughly the latitude of Kentucky and Virginia. Someone from Vermont, which is at 44 degrees, would be 0.025 less likely to migrate than someone at 37 degrees. The coefficients on all three variables suggest that individuals were less likely to migrate as the cost of doing so rose.

In column 3, when we add real wages and the share of foreign born in the state of birth in 1850, only age, literacy, distance, latitude, and latitude squared are significant. The coefficients on both average real wages for common laborers and the share foreign born are, however, of the expected sign. Individuals in states with higher average real wages for common laborers were less likely to migrate. Individuals in states with greater shares of foreign-born men are also more likely to migrate. This is consistent with a greater share of foreign born in a state negatively affecting labor markets and so playing a role in the decision to migrate.Footnote ³²

In columns 4 and 5, we include the weighted average of the characteristics of the state of last residence conditional on birthplace. For example, recall that for individuals born in New York, 46 percent came directly from New York, 16 percent came from Michigan, 10 percent came from Illinois, 8 percent came from Wisconsin, and the remaining 20 percent came via other states. The characteristics of the state of last residence would be the weighted average of New York, Michigan, Illinois, Wisconsin, and other states. The fit is improved relative to columns 2 and 3, suggesting that the characteristics of the state of last residence have more predictive power than the characteristics of birthplace do.

The most striking differences between columns 2 and 4 are that in column 4, the coefficients on latitude are of opposite sign and water transport is no longer statistically significant. Thus, conditional on distance from state of last residence, migrants are now slightly more likely to come from the periphery. This may reflect the fact that migrants are being disproportionately drawn from the upper Midwest and to a lesser degree from states such as Louisiana (see Figure 3). If we compare columns 3 and 5, we see that coefficients on latitude are also of opposite sign and the coefficient on foreign born is statistically significant. The latter is in line with a foreign-born push factor playing a role in the decision to migrate internally.Footnote ³³

Having examined who migrated, we now turn to how they did economically.

AVERAGE DAILY PRODUCT IN THE GOLD RUSH

The enumerator in El Dorado County recorded “average value of each miner's daily product” on the 1850 Census of Population.Footnote ³⁴ For miners who were mining on their own, average daily product would have corresponded to their earnings. For miners who were mining for others, their wage on average would have been less than the average daily product. Fifty-five percent of all men ages 20–40 reported positive average daily product from mining. Men who did not report values may have done so for at least four reasons. First, they may not have been mining. Second, they may have been unwilling to report values, if, for example, they thought that the information would be used for tax or other purposes. Third, they may have just arrived and so had no average daily product to report. This was particularly likely in the fall, when most overland immigrants arrived. Fourth, they may not have been present at the time the household was enumerated, and so value may not have been reported. We will investigate the issue of who reports average daily product in more detail shortly.

Table 5 presents the mean, standard deviation, and minimum and maximum reported average value of daily product for a number of groups, including all men, miners, native-born miners, and foreign-born miners. Although the range was relatively large, the median value was $4 and the twenty-fifth and seventy-fifth percentiles were $3 and $5 per day. These numbers are in line with data from a variety of contemporary sources.Footnote ³⁵ Margo, drawing on data from wages paid to civilian employees at U.S. Army forts in California during the gold rush, finds average daily wages of $3.78 per day in 1850.Footnote ³⁶ Data from the 1850 Census of Manufacturing for California on wages paid in mining also suggest that the average was close to $4 per day. Contemporary estimates were printed in the Sacramento Placer Times on 26 October 1850. The editor of that newspaper estimated that average daily earnings were Feather River, $6; Yuba River, $4; Bear River, $4; and American River $5.Footnote ³⁷ Another estimate from 1850 covering 14 companies engaged in river damming operations indicated that the average daily product was $3.16.Footnote ³⁸

Table 5

SUMMARY STATISTICS ON AVERAGE DAILY PRODUCT FROM MINING IN EL DORADO COUNTY IN 1850

Source:

See the text.

In Table 6, we present average daily product and the cost of board from the 1850 Social Statistics for El Dorado County, three other mining counties, a number of states, and the nation as a whole.Footnote ³⁹ These numbers indicate that the average miner in El Dorado who reported average daily product was, if he was working on his own account, earning somewhat less than what they would have earned as a day laborer in El Dorado County. This is not at all surprising, as miners were willing to accept a lower average payoff in the hopes that they would strike it rich.

To the extent that the weekly price of board reflected the actual cost of food, miners needed to spend roughly half their daily earnings on food. Some savings could probably be had by buying and preparing the food themselves, but the main point is that expenditures on food were a substantial share of earnings.

Was the average native-born miner earning more mining in 1850 than he would have had he not migrated? The answer will depend on his most recent state of residence and his previous occupation. Because it is useful for understanding the following results, and because it is important for results in the next section, in Table 7 we show the distribution of occupations for men ages 20–40 nationally and for selected states where many miners had last been resident. Nationally, the most common occupation for this age group was farmer, with 36 percent of men reporting that they were farmers. The next two most common occupations were craftsmen and laborers. Each had a 17 percent share.

Table 6

EARNINGS FROM SELECTED OCCUPATIONS IN 1850

Notes:

The county-level observations for California are from the manuscript schedules for Social Statistics for 1850, which are available through interlibrary loan. The state-level observations are from ICPSR 2896: Historical, Demographic, Economic, and Social Data: The United States, 1790–2000. The United States average is population weighted.

Because uniform statistics on the wages paid to day laborers were collected for all states as part of the Census of Social Statistics, we compare miners to day laborers. In absolute terms, miners who had previously been day laborers were certainly doing better. Table 6 showed the average state wages for a number of occupations nationally and for selected states where many miners had last been resident. If we multiply the U.S. average daily wage without board by six and subtract the weekly price of board, we find that the average day laborer would have made $3.26 per week after board. If we do the same exercise for El Dorado County (outside the cities), we find that the average day laborer made $22.00 per week after board. This amount is sufficiently large that it probably exceeded the earnings of most individuals in most occupations in the rest of the United States.

In relative terms, however, miners were also doing worse than a day laborer. If we use the cost of board as a price index, the average day laborer in the United States made 2.89 times the cost of board, whereas the average day laborer in El Dorado County made 2.57 times the cost of board. In comparison, the average miner in El Dorado County made 1.80 times the cost of board, which is substantially less than what the average day laborer in the United States made. To the extent that miners were drawn from more lucrative occupations than day laborer, they would have been doing even worse in relative terms.

Table 7

DISTRIBUTION OF OCCUPATIONS IN 1850 FOR MALES AGES 20–40

Notes:

Author's computations from the public use sample of the 1850 Census of Population.

In the first two columns of Table 8, we examine the question: Who reported positive average daily product from mining? We estimate a regression of the following form

(2)

The dependent variable is whether individual i reported positive daily average daily product from mining (one) or not (zero). This is regressed on a vector of the individual's personal characteristics (X_i). In column 1, we only include the characteristics that are clearly exogenous such as age, race, literacy, and whether the individual was foreign born and foreign born but from an English-speaking area. The explanatory power in column 1 is effectively zero, although the coefficients on literacy, foreign born, and foreign-born English speaking are all statistically significant.

In column 2, we add two additional characteristics: when the individual was enumerated and whether the individual listed his occupation as miner. The effects of both variables are substantial. Individuals who were enumerated later were more likely than individuals enumerated earlier to report average daily product from mining. This may well be the result of the enumerator starting the census in more populous locations, which had proportionately fewer miners, and moving on to more remote locations, where the proportion of miners was higher. Nonminers were 64 percent less likely than individuals who listed other occupations to report average daily product from mining.

Table 8

DETERMINANTS OF AVERAGE DAILY PRODUCT FROM MINING IN EL DORADO COUNTY IN 1850

* denotes statistical significance at the 10 percent level.

** denotes statistical significance at the 5 percent level.

*** denotes statistical significance at the 1 percent level.

Notes:

Standard errors are in parentheses. All standard errors are robust.

In the next two columns of Table 8, we examine the question: Conditional on having reported average daily product, what were its determinants? We estimate a regression similar to specification (2), where now the dependent variable is ln(average daily product + 1). As before, the dependent variable is regressed on a vector of the individual's personal characteristics. In column 3, as in column 1, the exogenous characteristics have virtually no explanatory power. This is what we would expect if average daily product were effectively random. Individuals who were literate reported somewhat higher average daily product, and whites and foreign-born individuals who were from English-speaking nations reported slightly somewhat lower average daily product, but the effects are small. Unless there was variation in hours worked or certain miners were confined to poorer quality locations, we would expect all miners to have had a similar chance at success.

In column 4, we add when the individual was enumerated and whether the individual listed his occupation as miner. The explanatory power remains low, although we do find that individuals who were enumerated later reported higher average daily product, as did miners. The higher average daily product reported by individuals who were enumerated later is consistent with the enumerator starting the census in more populous (lower average daily product) locations then moving to more remote (higher average daily product) locations. Individuals who listed their occupation as miner were probably devoting more of their time to mining and therefore had higher average daily product.

Strikingly, average daily product reported by foreign-born non-English-speaking men (the coefficient on foreign born) was no different than those of native-born men. Of course it is possible that this is a selection story, whereby high earning foreign-born non-English-speaking miners, most of whom were Mexican, are more likely than high earning native-born miners to report their earnings. Also, there is some discussion in the historical literature that the Mexican miners were previously miners in Mexico and so had greater experience and skill than other groups. We are unable to evaluate these possibilities, beyond the analysis we already presented on who reported earnings.

If selection was not a significant factor, and this is a big if, this result is striking for two reasons. First, it casts some doubt on claims by historians that foreign miners, particularly non-English-speaking foreign miners, were uniformly marginalized or otherwise discriminated against during the gold rush.Footnote ⁴⁰ Second, both wage data from California during the gold rush and wage data from later in the nineteenth century show that foreign-born men—particularly newly arrived foreign-born non-English-speaking men—had lower wages than native-born men.Footnote ⁴¹ One would infer that this is at least in part due to lower productivity. Yet we see no evidence that foreign-born non-English-speaking men had lower productivity in gold mining in El Dorado County in 1850. The two observations are not necessarily inconsistent—Mexican miners may have been as productive as American miners, while Mexican teamsters may have been less productive than American teamsters. What the observations highlight is how little we know about wages and productivity during the nineteenth century.

REAL ESTATE WEALTH IN THE GOLD RUSH

To understand the economic success of individuals in California, it is helpful to consider wealth in addition to average daily product from mining. The 1850 Census of Population asked individuals to report the gross value of their real estate holdings. Unfortunately, the gross value of real estate holdings is a less than perfect proxy for real estate wealth, because of the possibility that individuals had real estate-related debt. Largely for convenience, and following in the tradition of other scholars, we will refer to the reported value as real estate wealth. Interpreting the values reported in the census is complicated for a number of other reasons. First, it was almost certainly true that individuals who owned real estate were less likely to migrate. Second, those who owned real estate may have sold it to finance the trip. Third, if the holding was in California, prices were fluctuating daily, so the accuracy of the reported value was likely to be lower than in other parts of the United States. Fourth, much of the real estate that individuals reported owning in California may have represented ownership of use rights to mining claims. Thus, any differences between California and the rest of the United States have to be interpreted cautiously.

Given that we will be comparing men in California with men who stayed behind, it is worth discussing what we expect to find. It is entirely possible that we would find that California was a bonanza for those who went, or more likely that it was a bust. These outcomes will depend on the extent to which expectations induced too many or too few men to migrate. Suppose, however, that on average expectations were correct. The outcomes in California were highly skewed. Thus, measured wealth may be lower for those who migrated, particularly if we use a natural log transformation, as is common with wealth data. It may also be lower, because of the cost of the trip or because individuals in California chose to invest their wealth in something other than real estate.

Table 9 estimates the effects of being in California on real estate wealth. We estimate a regression similar to specification (2), where now the dependent variable is ln(real estate wealth + 1). The results are sensitive to what one takes as the counterfactual—the set of occupations that one views miners as likely to have been drawn from. Based on the coding of occupations in the public use sample and our coding using the same criteria, individuals can be classified into one of ten occupational categories. In column 1, we use the full sample and include occupational fixed effects. The coefficient on the variable in California is negative and significant. What it measures is the average wealth difference between individuals in California and individuals elsewhere, holding occupation constant. In particular, operatives, the occupational grouping that includes miners, in California are being compared to operatives elsewhere.

Table 9

THE EFFECT OF BEING IN CALIFORNIA ON REAL PROPERTY HOLDINGS IN 1850 (dependent variable: ln real property)

* denotes statistical significance at the 10 percent level.

** denotes statistical significance at the 5 percent level.

*** denotes statistical significance at the 1 percent level.

Notes:

All regressions are weighed by the person weights (100 or 101 for the public use sample outside of California, 160 or 161.6 for the public use sample inside California, and 1.6 for El Dorado, Sutter, and Yuba Counties). All standard errors are clustered on birthplace. Distance in the first stage regression is distance from place of last residence. Only native-born men are included in the sample.

In columns 2 and 3, we split the sample to separately consider individuals who were operatives, service workers, laborers, farm laborers, or farmers, and individuals who were professionals, managers, clerks, craftsmen, and salesmen. Thus, the sample in column 2 contains miners, other operatives, and occupations that miners in California were likely to have been drawn from. These occupations all typically required substantial amounts of manual labor, and so men in these categories would be relatively well suited to the physical demands of mining. We do not include occupational fixed effects, because historical evidence suggests miners were drawn from these groups, and so they collectively represent the relevant comparison group. The coefficient on the variable in California remains negative and significant.

In column 3, we consider professionals, managers, clerks, crafts-men, and salesmen. In this specification we include occupational dummies, because individuals in California in these occupations were likely to have engaged in these occupations previously. So, for example, the relevant comparison group for professionals in California was professionals elsewhere. The coefficient on the variable in California is negative but smaller and statistically insignificant, suggesting that any negative effect on wealth was largely confined to the occupations in column 2.

One possible concern with what we have done so far concerns the inclusion of farmers in column 2. There are two important issues with respect to farmers. First, the individuals most likely to migrate to California were relatively unlikely to be farm owners. Instead, they may have been the children or relatives of farmers who were listed as farmers but did not actually own farms.Footnote ⁴² These individuals were arguably more similar to farm laborers than they were to farmers. Evidence from diaries indicates that a relatively small fraction of migrants were previously farmers.Footnote ⁴³ Second, farmers in California were much wealthier than their counterparts in the United States. Further, a disproportionate number of the individuals classified as farmers in California arrived prior to the gold rush.

In column 4, we run the same regression as in column 2, but now omit farmers from the comparison group for miners. The coefficient on the variable in California is positive, small, and statistically insignificant. In column 5, we run the same regression as in column 3, but now add farmers to this category. As was the case in column 3, the coefficient on the variable in California is negative and statistically insignificant.

Thus far, we have not discussed selection effects, although it is clear that different types of men were probably migrating from Missouri than from Connecticut. In the bottom panel of Table 9, we instrument for the decision to migrate using the distance from the state of last residence to California. As we saw in Table 4, where we used a logit to examine the determinants of migration to California, other characteristics of the state of last residence were sometimes statistically significant. Their inclusion tended, however, to depress the F-statistic in the first stage regression.

Once we control for selection in the bottom panel of Table 9, the effect of being in California is now generally positive and ranges from very large positive and significant to small and insignificant. In column 1, the average effect of being in California on those who went is positive, statistically significant, and large. In columns 2 and 4, where we focus on miners and occupations that they were likely to have come from, we see that the results are very sensitive to the inclusion of farmers. When we include farmers in the set of occupations that miners would be drawn from, the effect is positive, statistically significant, and large. When we exclude farmers from the set of occupations that miners would be drawn from, the effect is negative, statistically insignificant, and small. We believe the latter specification is preferable given the problems with farmers we outlined above. In columns 3 and 5, we examine the success of nonmining occupations in California, first with farmers excluded and then with farmers included. The effect is large and statistically significant in column 3, when farmers are excluded, and large but not statistically significant in column 5, when farmers are included. Because of the problems with farmers, we prefer the former specification.

Did the average native-born man in California in 1850 have greater real estate holdings than he would have had he not migrated? The answer appears to have depended very much on his occupation both before and after migrating. To the extent that miners in California were drawn from the ranks of operatives, service workers, laborers, and farm laborers (column 4), the effect of migrating on their wealth was effectively zero. To the extent that nonminers in California were previously professionals, managers, clerks, craftsmen, and salesmen, the effect of migrating on their wealth was positive and large (column 3). Interestingly, these results are in line with commentary during the gold rush and with later historical analysis. Both have indicated that the bulk of the wealth earned during the gold rush was earned by merchants, proprietors of hotels, doctors, and other service providers. The returns to mining were low, because of the large numbers of migrants and the ease of entry into mining.